Hi Al,
I think this patch should include all the fixes necessary. I could merge
it in, but I think it might be better to tag it on the end as an additional
patch.
David
---
iov_iter: Four fixes for ITER_XARRAY
Fix four things[1] in the patch that adds ITER_XARRAY[2]:
(1) Remove the address_space struct predeclaration. This is a holdover
from when it was ITER_MAPPING.
(2) Fix _copy_mc_to_iter() so that the xarray segment updates count and
iov_offset in the iterator before returning.
(3) Fix iov_iter_alignment() to not loop in the xarray case. Because the
middle pages are all whole pages, only the end pages need be
considered - and this can be reduced to just looking at the start
position in the xarray and the iteration size.
(4) Fix iov_iter_advance() to limit the size of the advance to no more
than the remaining iteration size.
Reported-by: Al Viro <[email protected]>
Signed-off-by: David Howells <[email protected]>
Link: https://lore.kernel.org/r/[email protected] [1]
Link: https://lore.kernel.org/r/161918448151.3145707.11541538916600921083.stgit@warthog.procyon.org.uk [2]
---
include/linux/uio.h | 1 -
lib/iov_iter.c | 5 +++++
2 files changed, 5 insertions(+), 1 deletion(-)
diff --git a/include/linux/uio.h b/include/linux/uio.h
index 5f5ffc45d4aa..d3ec87706d75 100644
--- a/include/linux/uio.h
+++ b/include/linux/uio.h
@@ -10,7 +10,6 @@
#include <uapi/linux/uio.h>
struct page;
-struct address_space;
struct pipe_inode_info;
struct kvec {
diff --git a/lib/iov_iter.c b/lib/iov_iter.c
index 44fa726a8323..61228a6c69f8 100644
--- a/lib/iov_iter.c
+++ b/lib/iov_iter.c
@@ -791,6 +791,8 @@ size_t _copy_mc_to_iter(const void *addr, size_t bytes, struct iov_iter *i)
curr_addr = (unsigned long) from;
bytes = curr_addr - s_addr - rem;
rcu_read_unlock();
+ i->iov_offset += bytes;
+ i->count -= bytes;
return bytes;
}
})
@@ -1147,6 +1149,7 @@ void iov_iter_advance(struct iov_iter *i, size_t size)
return;
}
if (unlikely(iov_iter_is_xarray(i))) {
+ size = min(size, i->count);
i->iov_offset += size;
i->count -= size;
return;
@@ -1346,6 +1349,8 @@ unsigned long iov_iter_alignment(const struct iov_iter *i)
return size | i->iov_offset;
return size;
}
+ if (unlikely(iov_iter_is_xarray(i)))
+ return (i->xarray_start + i->iov_offset) | i->count;
iterate_all_kinds(i, size, v,
(res |= (unsigned long)v.iov_base | v.iov_len, 0),
res |= v.bv_offset | v.bv_len,
On Mon, Apr 26, 2021 at 12:14:50AM +0100, David Howells wrote:
> Hi Al,
>
> I think this patch should include all the fixes necessary. I could merge
> it in, but I think it might be better to tag it on the end as an additional
> patch.
Looks sane, but I wonder if it would've been better to deal with this
> @@ -791,6 +791,8 @@ size_t _copy_mc_to_iter(const void *addr, size_t bytes, struct iov_iter *i)
> curr_addr = (unsigned long) from;
> bytes = curr_addr - s_addr - rem;
> rcu_read_unlock();
> + i->iov_offset += bytes;
> + i->count -= bytes;
> return bytes;
> }
> })
by having your iterator check the return value of X callback and, having
decremented .bv_len by return value, broke out of the loop.
__label__ __bugger_off;
xas_for_each(&xas, head, ULONG_MAX) { \
if (xas_retry(&xas, head)) \
continue; \
if (WARN_ON(xa_is_value(head))) \
break; \
if (WARN_ON(PageHuge(head))) \
break; \
for (j = (head->index < index) ? index - head->index : 0; \
j < thp_nr_pages(head); j++) { \
__v.bv_page = head + j; \
size_t left;
offset = (i->xarray_start + skip) & ~PAGE_MASK; \
seg = PAGE_SIZE - offset; \
__v.bv_offset = offset; \
__v.bv_len = min(n, seg); \
left = (STEP);
__v.bv_len -= left;
n -= __v.bv_len; \
skip += __v.bv_len; \
if (!n || left)
goto __bugger_off;
} \
if (n == 0) \
break; \
} \
__bugger_off:
Then rename iterate_and_advance() to __iterate_and_advance() and have
#define iterate_and_advance(....., X) __iterate_and_advance(....., ((void)(X),0))
with iterate_all_kinds() using iterate_xarray(....,((void)(X),0)
Then _copy_mc_to_iter() could use __iterate_and_advance(), getting rid of
the need of doing anything special in case of short copy. OTOH, I can do
that myself in a followup - not a problem.
On Mon, 2021-04-26 at 00:14 +0100, David Howells wrote:
> Hi Al,
>
> I think this patch should include all the fixes necessary. I could merge
> it in, but I think it might be better to tag it on the end as an additional
> patch.
>
> David
> ---
> iov_iter: Four fixes for ITER_XARRAY
>
> Fix four things[1] in the patch that adds ITER_XARRAY[2]:
>
> (1) Remove the address_space struct predeclaration. This is a holdover
> from when it was ITER_MAPPING.
>
> (2) Fix _copy_mc_to_iter() so that the xarray segment updates count and
> iov_offset in the iterator before returning.
>
> (3) Fix iov_iter_alignment() to not loop in the xarray case. Because the
> middle pages are all whole pages, only the end pages need be
> considered - and this can be reduced to just looking at the start
> position in the xarray and the iteration size.
>
> (4) Fix iov_iter_advance() to limit the size of the advance to no more
> than the remaining iteration size.
>
> Reported-by: Al Viro <[email protected]>
> Signed-off-by: David Howells <[email protected]>
> Link: https://lore.kernel.org/r/[email protected] [1]
> Link: https://lore.kernel.org/r/161918448151.3145707.11541538916600921083.stgit@warthog.procyon.org.uk [2]
> ---
> include/linux/uio.h | 1 -
> lib/iov_iter.c | 5 +++++
> 2 files changed, 5 insertions(+), 1 deletion(-)
>
> diff --git a/include/linux/uio.h b/include/linux/uio.h
> index 5f5ffc45d4aa..d3ec87706d75 100644
> --- a/include/linux/uio.h
> +++ b/include/linux/uio.h
> @@ -10,7 +10,6 @@
> #include <uapi/linux/uio.h>
>
>
>
>
> struct page;
> -struct address_space;
> struct pipe_inode_info;
>
>
>
>
> struct kvec {
> diff --git a/lib/iov_iter.c b/lib/iov_iter.c
> index 44fa726a8323..61228a6c69f8 100644
> --- a/lib/iov_iter.c
> +++ b/lib/iov_iter.c
> @@ -791,6 +791,8 @@ size_t _copy_mc_to_iter(const void *addr, size_t bytes, struct iov_iter *i)
> curr_addr = (unsigned long) from;
> bytes = curr_addr - s_addr - rem;
> rcu_read_unlock();
> + i->iov_offset += bytes;
> + i->count -= bytes;
> return bytes;
> }
> })
> @@ -1147,6 +1149,7 @@ void iov_iter_advance(struct iov_iter *i, size_t size)
> return;
> }
> if (unlikely(iov_iter_is_xarray(i))) {
> + size = min(size, i->count);
> i->iov_offset += size;
> i->count -= size;
> return;
> @@ -1346,6 +1349,8 @@ unsigned long iov_iter_alignment(const struct iov_iter *i)
> return size | i->iov_offset;
> return size;
> }
> + if (unlikely(iov_iter_is_xarray(i)))
> + return (i->xarray_start + i->iov_offset) | i->count;
> iterate_all_kinds(i, size, v,
> (res |= (unsigned long)v.iov_base | v.iov_len, 0),
> res |= v.bv_offset | v.bv_len,
>
I did a test run with your v7 pile, this patch, and my ceph fscache
rework patches and it did fine. You can add:
Tested-by: Jeff Layton <[email protected]>
On Sun, Apr 25, 2021 at 7:15 PM David Howells <[email protected]> wrote:
>
> Hi Al,
>
> I think this patch should include all the fixes necessary. I could merge
> it in, but I think it might be better to tag it on the end as an additional
> patch.
>
> David
> ---
> iov_iter: Four fixes for ITER_XARRAY
>
> Fix four things[1] in the patch that adds ITER_XARRAY[2]:
>
> (1) Remove the address_space struct predeclaration. This is a holdover
> from when it was ITER_MAPPING.
>
> (2) Fix _copy_mc_to_iter() so that the xarray segment updates count and
> iov_offset in the iterator before returning.
>
> (3) Fix iov_iter_alignment() to not loop in the xarray case. Because the
> middle pages are all whole pages, only the end pages need be
> considered - and this can be reduced to just looking at the start
> position in the xarray and the iteration size.
>
> (4) Fix iov_iter_advance() to limit the size of the advance to no more
> than the remaining iteration size.
>
> Reported-by: Al Viro <[email protected]>
> Signed-off-by: David Howells <[email protected]>
> Link: https://lore.kernel.org/r/[email protected] [1]
> Link: https://lore.kernel.org/r/161918448151.3145707.11541538916600921083.stgit@warthog.procyon.org.uk [2]
> ---
> include/linux/uio.h | 1 -
> lib/iov_iter.c | 5 +++++
> 2 files changed, 5 insertions(+), 1 deletion(-)
>
> diff --git a/include/linux/uio.h b/include/linux/uio.h
> index 5f5ffc45d4aa..d3ec87706d75 100644
> --- a/include/linux/uio.h
> +++ b/include/linux/uio.h
> @@ -10,7 +10,6 @@
> #include <uapi/linux/uio.h>
>
> struct page;
> -struct address_space;
> struct pipe_inode_info;
>
> struct kvec {
> diff --git a/lib/iov_iter.c b/lib/iov_iter.c
> index 44fa726a8323..61228a6c69f8 100644
> --- a/lib/iov_iter.c
> +++ b/lib/iov_iter.c
> @@ -791,6 +791,8 @@ size_t _copy_mc_to_iter(const void *addr, size_t bytes, struct iov_iter *i)
> curr_addr = (unsigned long) from;
> bytes = curr_addr - s_addr - rem;
> rcu_read_unlock();
> + i->iov_offset += bytes;
> + i->count -= bytes;
> return bytes;
> }
> })
> @@ -1147,6 +1149,7 @@ void iov_iter_advance(struct iov_iter *i, size_t size)
> return;
> }
> if (unlikely(iov_iter_is_xarray(i))) {
> + size = min(size, i->count);
> i->iov_offset += size;
> i->count -= size;
> return;
> @@ -1346,6 +1349,8 @@ unsigned long iov_iter_alignment(const struct iov_iter *i)
> return size | i->iov_offset;
> return size;
> }
> + if (unlikely(iov_iter_is_xarray(i)))
> + return (i->xarray_start + i->iov_offset) | i->count;
> iterate_all_kinds(i, size, v,
> (res |= (unsigned long)v.iov_base | v.iov_len, 0),
> res |= v.bv_offset | v.bv_len,
>
You can add
Tested-by: Dave Wysochanski <[email protected]>
I added this patch on top of your v7 series then added my current
NFS patches to use netfs lib.
I ran xfstests with fscache enabled on NFS versions (3, 4.0, 4.1, 4.2),
as well as connectathon and some unit tests.